platform/x86/intel/ifs: Add an entry rendezvous for SAF

The activation for Scan at Field (SAF) includes a parameter to make
microcode wait for both threads to join. It's preferable to perform an
entry rendezvous before the activation to ensure that they start the
`wrmsr` close enough to each other. In some cases it has been observed
that one of the threads might be just a bit late to arrive. An entry
rendezvous reduces the likelihood of these cases occurring.

Add an entry rendezvous to ensure the activation on both threads happen
close enough to each other.

Signed-off-by: Ashok Raj <ashok.raj@intel.com>
Reviewed-by: Tony Luck <tony.luck@intel.com>
Link: https://lore.kernel.org/r/20240125082254.424859-6-ashok.raj@intel.com
Reviewed-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>

1 files changed, 26 insertions, 22 deletions

diff --git a/drivers/platform/x86/intel/ifs/runtest.c b/drivers/platform/x86/intel/ifs/runtest.c
index e3307dd8e3c4..95b4b71fab53 100644
--- a/drivers/platform/x86/intel/ifs/runtest.c
+++ b/drivers/platform/x86/intel/ifs/runtest.c
@@ -140,6 +140,29 @@ static bool can_restart(union ifs_status status)
 	return false;
 }
 
+#define SPINUNIT 100 /* 100 nsec */
+static atomic_t array_cpus_in;
+static atomic_t scan_cpus_in;
+
+/*
+ * Simplified cpu sibling rendezvous loop based on microcode loader __wait_for_cpus()
+ */
+static void wait_for_sibling_cpu(atomic_t *t, long long timeout)
+{
+	int cpu = smp_processor_id();
+	const struct cpumask *smt_mask = cpu_smt_mask(cpu);
+	int all_cpus = cpumask_weight(smt_mask);
+
+	atomic_inc(t);
+	while (atomic_read(t) < all_cpus) {
+		if (timeout < SPINUNIT)
+			return;
+		ndelay(SPINUNIT);
+		timeout -= SPINUNIT;
+		touch_nmi_watchdog();
+	}
+}
+
 /*
  * Execute the scan. Called "simultaneously" on all threads of a core
  * at high priority using the stop_cpus mechanism.
@@ -165,6 +188,8 @@ static int doscan(void *data)
 	/* Only the first logical CPU on a core reports result */
 	first = cpumask_first(cpu_smt_mask(cpu));
 
+	wait_for_sibling_cpu(&scan_cpus_in, NSEC_PER_SEC);
+
 	/*
 	 * This WRMSR will wait for other HT threads to also write
 	 * to this MSR (at most for activate.delay cycles). Then it
@@ -230,6 +255,7 @@ static void ifs_test_core(int cpu, struct device *dev)
 		}
 
 		params.activate = &activate;
+		atomic_set(&scan_cpus_in, 0);
 		stop_core_cpuslocked(cpu, doscan, &params);
 
 		status = params.status;
@@ -270,28 +296,6 @@ static void ifs_test_core(int cpu, struct device *dev)
 	}
 }
 
-#define SPINUNIT 100 /* 100 nsec */
-static atomic_t array_cpus_in;
-
-/*
- * Simplified cpu sibling rendezvous loop based on microcode loader __wait_for_cpus()
- */
-static void wait_for_sibling_cpu(atomic_t *t, long long timeout)
-{
-	int cpu = smp_processor_id();
-	const struct cpumask *smt_mask = cpu_smt_mask(cpu);
-	int all_cpus = cpumask_weight(smt_mask);
-
-	atomic_inc(t);
-	while (atomic_read(t) < all_cpus) {
-		if (timeout < SPINUNIT)
-			return;
-		ndelay(SPINUNIT);
-		timeout -= SPINUNIT;
-		touch_nmi_watchdog();
-	}
-}
-
 static int do_array_test(void *data)
 {
 	union ifs_array *command = data;